Some change in CasePath behaviour noticed

[rehsals]

I've been using TCA version 0.19.0 and today I've updated to 0.22.0.
And instantly I've got several breakpoints about nil-states when corresponding reducer gets an action.

My setup was as follows:

struct LocalState {}
enum LocalAction {
  case whatever
}

let localReducer: Reducer<LocalState, LocalAction, Void> = <...>

struct GlobalState {
  enum Option {
    case local(LocalState)
  }

  var option: Option? <--- optionality is important I guess
}

enum GlobalAction {
  case local(LocalAction)
}

let globalReducer = Reducer<GlobalState, GlobalAction, Void>.combine(
  localReducer
    .pullback(
            state: /GlobalState.Option.local,
            action: .`self`,
            environment: { $0 })
    .pullback(
            state: \.option,
            action: /GlobalAction.local,
            environment: { $0 }),
  .init { state, action, _ in <...> }
)

So CasePath from GlobalState.Option? to LocalState worked fine on 0.19.0 and not working anymore.
I've changed my code to be:

let globalReducer = Reducer<GlobalState, GlobalAction, Void>.combine(
  localReducer
    .pullback(
            state: /GlobalState.Option.local,
            action: .`self`,
            environment: { $0 })
    .pullback(
            state: /Optional<GlobalState.Option>.some,
            action: .`self`,
            environment: { $0 })
    .pullback(
            state: \.option,
            action: /GlobalAction.local,
            environment: { $0 }),
  .init { state, action, _ in <...> }
)

and things got back to normal.
Is this is an expected behaviour and a bug being fixed? May be it could be useful for someone in similar situations

Edit: probably it's a CasePath version what is more relevant. My current is 0.5.0. Unfortunately I can't tell which version I was using before the update
Edit2: there is an issue in CasePath repo which seems related: pointfreeco/swift-case-paths#40

[stephencelis]

@rehsals This is indeed pointfreeco/swift-case-paths#40. Swift's optional promotion can promote embed functions to optional-returning embed functions, which breaks some of the expectations made in the new runtime code. We hope to workaround this problem when we have time, but in the meantime using an explicit /Optional.some is definitely the way to go.

One thing you can do, though, is avoid an extra pullback by using CasePath.appending(path:) instead:

    .pullback(
            state: (/Optional.some).appending(path: /GlobalState.Option.local),
            action: .`self`,
            environment: { $0 })

[rehsals]

Got it. Nice hint for avoiding extra pullback, thanks!